NVIDIA Model Optimizer is a library that changes deep-learning models through techniques such as quantization, pruning, distillation, neural architecture search, speculative decoding, and sparsity so they can run more efficiently. Developers use it to optimize Hugging Face, PyTorch, or ONNX models and export checkpoints for inference frameworks such as SGLang, TensorRT-LLM, TensorRT, or vLLM. The catalogue entries provide skills, instructions, and a plugin for using its workflows.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add NVIDIA/Model-Optimizer --skill release-cherry-pickgit clone --depth 1 https://github.com/NVIDIA/Model-OptimizerWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/nvidia/model-optimizer/release-cherry-pick)<a href="https://agentmods.dev/skills/nvidia/model-optimizer/release-cherry-pick"><img src="https://agentmods.dev/badge/skills/nvidia/model-optimizer/release-cherry-pick/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/nvidia/model-optimizer/release-cherry-pick"><img src="https://agentmods.dev/badge/skills/nvidia/model-optimizer/release-cherry-pick.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00083 | $0.01880 |
| Opus 5.5 | $0.00033 | $0.00752 |
| Sonnet 5.5 | $0.00017 | $0.00376 |
| Haiku 4.5 | $0.00008 | $0.00188 |
Grade A, and why
release-cherry-pick scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 13d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 180 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Cherry-pick PRs to a Release Branch
Cherry-pick all merged main PRs labeled cherry-pick-<version> (but not cherry-pick-done) into the corresponding release/<version> branch, one by one in merge order.
Step 1 — Identify the target version
Ask the user for the release version (e.g. 0.44.0) and target release candidate (e.g. rc2) if not already provided. Confirm whether this is the final RC.
Set VERSION=<version> for use in subsequent steps.
Step 2 — Audit candidates for missing labels
Before fetching the labeled queue, audit for missing bug fixes:
- Non-final RC: audit release NVBugs and recent merged PRs.
- Final RC: audit only fixes tied to NVBugs with the exact keyword
Committed_ModelOpt_<VERSION>and QA-verification evidence for the fix. Verify that evidence in the NVBug or linked QA results before recommending Needs label; if missing or unclear, classify as Unknown and ask for confirmation. Skip the general recent-PR audit; do not introduce general bug-fix PRs or fixes tied only to NVBugs committed for another release.
Audit release NVBugs
Use the NVBugs MCP to search for every NVBug whose Keywords field contains the exact keyword Committed_ModelOpt_<VERSION>. Follow pagination until no next token is returned.
Fetch each matching NVBug with its comments. Extract every Model-Optimizer PR link or unambiguous PR #<NUM> reference from the comments, not only the latest comment. Keep only PRs that are merged into main, lack cherry-pick-<VERSION>, and are not already included in the release branch. Verify each remaining PR is a bug fix. The NVBug keyword is evidence for review, not by itself proof that every linked PR should be picked.
Audit recent merged PRs
For non-final RCs only: Audit PRs merged into main since the latest release candidate. Exclude PRs that already have cherry-pick-<VERSION> and changes already present on the release branch:
git fetch origin main "release/$VERSION" --tags
RC_TAG=$(git tag --merged "origin/release/$VERSION" \
--list "${VERSION}rc*" --sort=-version:refname | head -1)
test -n "$RC_TAG" || {
echo "No ${VERSION}rc* tag found on origin/release/$VERSION" >&2
exit 1
}
SINCE=$(git for-each-ref --format='%(creatordate:iso-strict)' \
"refs/tags/$RC_TAG")
PATCH_STATUS=$(git cherry "origin/release/$VERSION" origin/main)
SEARCH_RESULTS=$(
gh search prs \
--repo NVIDIA/Model-Optimizer \
--merged \
--base main \
--merged-at ">=$SINCE" \
--limit 1000 \
--json number,title,author,labels,url \
-- "-label:cherry-pick-$VERSION"
)
if test "$(jq 'length' <<<"$SEARCH_RESULTS")" -ge 1000; then
echo "Recent-PR audit reached the 1,000-result limit" >&2
exit 1
fi
jq -r '.[].number' <<<"$SEARCH_RESULTS" \
| while read -r pr; do
sha=$(gh pr view "$pr" --repo NVIDIA/Model-Optimizer \
--json mergeCommit --jq '.mergeCommit.oid')
if grep -q "^+ $sha$" <<<"$PATCH_STATUS"; then
gh pr view "$pr" --repo NVIDIA/Model-Optimizer \
--json number,title,author,labels,url
fi
done
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 13d ago Changed · +3 lines c1f6555cb5bb
- 1mo ago Changed · +87 lines · +13 tokens per session c32e2b70fc79
- 1mo ago First seen · 90 lines · 70 tokens per session scan A 47fab972cc05
release-cherry-pick is a skill published in the GitHub repository NVIDIA/Model-Optimizer (5,398 stars, last pushed today), licensed Apache-2.0. It adds 83 tokens to every session and 1,880 once invoked, about $0.0003 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
release-announcement
Write a release announcement — changelog, blog post, in-app note, or social post — that leads with user impact, names the audience, and includes upgrade/migration steps without filler.
release
Fully automate a release with version-bump and childhood.
release-notes
Generate user-facing release notes from tickets, PRDs, or changelogs. Creates clear, engaging summaries organized by category (new features, improvements, fixes). Use when writing release notes, creating changelogs, announcing product updates, or summarizing what shipped.
multi-agent-release-manager
Cleans up the workspace, formats code, runs presubmit checks, and uploads CLs to Gerrit.
pack-submit
Package one of this agent's own skills as a standalone community pack and submit it to the aeon registry as a PR.
release-helper
Ensures configuration and code changes are released correctly. Use PROACTIVELY whenever you edit config.json, change any setting, or fix a configuration bug in this project, so the change reaches production.